ib_srp.c 57.5 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39
/*
 * Copyright (c) 2005 Cisco Systems.  All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

#include <linux/module.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/err.h>
#include <linux/string.h>
#include <linux/parser.h>
#include <linux/random.h>
40
#include <linux/jiffies.h>
41 42 43 44 45 46 47

#include <asm/atomic.h>

#include <scsi/scsi.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_dbg.h>
#include <scsi/srp.h>
48
#include <scsi/scsi_transport_srp.h>
49 50 51 52 53 54 55 56 57 58 59 60 61

#include "ib_srp.h"

#define DRV_NAME	"ib_srp"
#define PFX		DRV_NAME ": "
#define DRV_VERSION	"0.2"
#define DRV_RELDATE	"November 1, 2005"

MODULE_AUTHOR("Roland Dreier");
MODULE_DESCRIPTION("InfiniBand SCSI RDMA Protocol initiator "
		   "v" DRV_VERSION " (" DRV_RELDATE ")");
MODULE_LICENSE("Dual BSD/GPL");

62 63 64 65 66
static int srp_sg_tablesize = SRP_DEF_SG_TABLESIZE;
static int srp_max_iu_len;

module_param(srp_sg_tablesize, int, 0444);
MODULE_PARM_DESC(srp_sg_tablesize,
67
		 "Max number of gather/scatter entries per I/O (default is 12, max 255)");
68

69 70 71 72 73 74
static int topspin_workarounds = 1;

module_param(topspin_workarounds, int, 0444);
MODULE_PARM_DESC(topspin_workarounds,
		 "Enable workarounds for Topspin/Cisco SRP target bugs if != 0");

75 76 77 78 79 80
static int mellanox_workarounds = 1;

module_param(mellanox_workarounds, int, 0444);
MODULE_PARM_DESC(mellanox_workarounds,
		 "Enable workarounds for Mellanox SRP target bugs if != 0");

81 82
static void srp_add_one(struct ib_device *device);
static void srp_remove_one(struct ib_device *device);
83 84
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr);
static void srp_send_completion(struct ib_cq *cq, void *target_ptr);
85 86
static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event);

87 88
static struct scsi_transport_template *ib_srp_transport_template;

89 90 91 92 93 94
static struct ib_client srp_client = {
	.name   = "srp",
	.add    = srp_add_one,
	.remove = srp_remove_one
};

95 96
static struct ib_sa_client srp_sa_client;

97 98 99 100 101 102 103 104 105 106
static inline struct srp_target_port *host_to_target(struct Scsi_Host *host)
{
	return (struct srp_target_port *) host->hostdata;
}

static const char *srp_target_info(struct Scsi_Host *host)
{
	return host_to_target(host)->target_name;
}

107 108 109
static int srp_target_is_topspin(struct srp_target_port *target)
{
	static const u8 topspin_oui[3] = { 0x00, 0x05, 0xad };
110
	static const u8 cisco_oui[3]   = { 0x00, 0x1b, 0x0d };
111 112

	return topspin_workarounds &&
113 114
		(!memcmp(&target->ioc_guid, topspin_oui, sizeof topspin_oui) ||
		 !memcmp(&target->ioc_guid, cisco_oui, sizeof cisco_oui));
115 116 117 118 119 120 121 122 123 124
}

static int srp_target_is_mellanox(struct srp_target_port *target)
{
	static const u8 mellanox_oui[3] = { 0x00, 0x02, 0xc9 };

	return mellanox_workarounds &&
		!memcmp(&target->ioc_guid, mellanox_oui, sizeof mellanox_oui);
}

125 126 127 128 129 130 131 132 133 134 135 136 137 138
static struct srp_iu *srp_alloc_iu(struct srp_host *host, size_t size,
				   gfp_t gfp_mask,
				   enum dma_data_direction direction)
{
	struct srp_iu *iu;

	iu = kmalloc(sizeof *iu, gfp_mask);
	if (!iu)
		goto out;

	iu->buf = kzalloc(size, gfp_mask);
	if (!iu->buf)
		goto out_free_iu;

139 140 141
	iu->dma = ib_dma_map_single(host->srp_dev->dev, iu->buf, size,
				    direction);
	if (ib_dma_mapping_error(host->srp_dev->dev, iu->dma))
142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161
		goto out_free_buf;

	iu->size      = size;
	iu->direction = direction;

	return iu;

out_free_buf:
	kfree(iu->buf);
out_free_iu:
	kfree(iu);
out:
	return NULL;
}

static void srp_free_iu(struct srp_host *host, struct srp_iu *iu)
{
	if (!iu)
		return;

162 163
	ib_dma_unmap_single(host->srp_dev->dev, iu->dma, iu->size,
			    iu->direction);
164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182
	kfree(iu->buf);
	kfree(iu);
}

static void srp_qp_event(struct ib_event *event, void *context)
{
	printk(KERN_ERR PFX "QP event %d\n", event->event);
}

static int srp_init_qp(struct srp_target_port *target,
		       struct ib_qp *qp)
{
	struct ib_qp_attr *attr;
	int ret;

	attr = kmalloc(sizeof *attr, GFP_KERNEL);
	if (!attr)
		return -ENOMEM;

183 184 185 186
	ret = ib_find_pkey(target->srp_host->srp_dev->dev,
			   target->srp_host->port,
			   be16_to_cpu(target->path.pkey),
			   &attr->pkey_index);
187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
	if (ret)
		goto out;

	attr->qp_state        = IB_QPS_INIT;
	attr->qp_access_flags = (IB_ACCESS_REMOTE_READ |
				    IB_ACCESS_REMOTE_WRITE);
	attr->port_num        = target->srp_host->port;

	ret = ib_modify_qp(qp, attr,
			   IB_QP_STATE		|
			   IB_QP_PKEY_INDEX	|
			   IB_QP_ACCESS_FLAGS	|
			   IB_QP_PORT);

out:
	kfree(attr);
	return ret;
}

D
David Dillow 已提交
206 207 208 209
static int srp_new_cm_id(struct srp_target_port *target)
{
	struct ib_cm_id *new_cm_id;

210
	new_cm_id = ib_create_cm_id(target->srp_host->srp_dev->dev,
D
David Dillow 已提交
211 212 213 214 215 216 217 218 219 220 221
				    srp_cm_handler, target);
	if (IS_ERR(new_cm_id))
		return PTR_ERR(new_cm_id);

	if (target->cm_id)
		ib_destroy_cm_id(target->cm_id);
	target->cm_id = new_cm_id;

	return 0;
}

222 223 224 225 226 227 228 229 230
static int srp_create_target_ib(struct srp_target_port *target)
{
	struct ib_qp_init_attr *init_attr;
	int ret;

	init_attr = kzalloc(sizeof *init_attr, GFP_KERNEL);
	if (!init_attr)
		return -ENOMEM;

231 232 233 234
	target->recv_cq = ib_create_cq(target->srp_host->srp_dev->dev,
				       srp_recv_completion, NULL, target, SRP_RQ_SIZE, 0);
	if (IS_ERR(target->recv_cq)) {
		ret = PTR_ERR(target->recv_cq);
235
		goto err;
236 237
	}

238 239 240 241
	target->send_cq = ib_create_cq(target->srp_host->srp_dev->dev,
				       srp_send_completion, NULL, target, SRP_SQ_SIZE, 0);
	if (IS_ERR(target->send_cq)) {
		ret = PTR_ERR(target->send_cq);
242
		goto err_recv_cq;
243 244 245
	}

	ib_req_notify_cq(target->recv_cq, IB_CQ_NEXT_COMP);
246 247 248 249 250 251 252 253

	init_attr->event_handler       = srp_qp_event;
	init_attr->cap.max_send_wr     = SRP_SQ_SIZE;
	init_attr->cap.max_recv_wr     = SRP_RQ_SIZE;
	init_attr->cap.max_recv_sge    = 1;
	init_attr->cap.max_send_sge    = 1;
	init_attr->sq_sig_type         = IB_SIGNAL_ALL_WR;
	init_attr->qp_type             = IB_QPT_RC;
254 255
	init_attr->send_cq             = target->send_cq;
	init_attr->recv_cq             = target->recv_cq;
256

257
	target->qp = ib_create_qp(target->srp_host->srp_dev->pd, init_attr);
258 259
	if (IS_ERR(target->qp)) {
		ret = PTR_ERR(target->qp);
260
		goto err_send_cq;
261 262 263
	}

	ret = srp_init_qp(target, target->qp);
264 265
	if (ret)
		goto err_qp;
266

267 268 269 270 271 272 273 274 275 276 277 278 279
	kfree(init_attr);
	return 0;

err_qp:
	ib_destroy_qp(target->qp);

err_send_cq:
	ib_destroy_cq(target->send_cq);

err_recv_cq:
	ib_destroy_cq(target->recv_cq);

err:
280 281 282 283 284 285 286 287 288
	kfree(init_attr);
	return ret;
}

static void srp_free_target_ib(struct srp_target_port *target)
{
	int i;

	ib_destroy_qp(target->qp);
289 290
	ib_destroy_cq(target->send_cq);
	ib_destroy_cq(target->recv_cq);
291 292 293

	for (i = 0; i < SRP_RQ_SIZE; ++i)
		srp_free_iu(target->srp_host, target->rx_ring[i]);
294
	for (i = 0; i < SRP_SQ_SIZE; ++i)
295 296 297 298 299 300 301 302 303 304 305
		srp_free_iu(target->srp_host, target->tx_ring[i]);
}

static void srp_path_rec_completion(int status,
				    struct ib_sa_path_rec *pathrec,
				    void *target_ptr)
{
	struct srp_target_port *target = target_ptr;

	target->status = status;
	if (status)
306 307
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Got failed path rec status %d\n", status);
308 309 310 311 312 313 314 315 316 317 318
	else
		target->path = *pathrec;
	complete(&target->done);
}

static int srp_lookup_path(struct srp_target_port *target)
{
	target->path.numb_path = 1;

	init_completion(&target->done);

319
	target->path_query_id = ib_sa_path_rec_get(&srp_sa_client,
320
						   target->srp_host->srp_dev->dev,
321 322
						   target->srp_host->port,
						   &target->path,
323
						   IB_SA_PATH_REC_SERVICE_ID	|
324 325 326 327 328 329 330 331 332 333 334 335 336 337
						   IB_SA_PATH_REC_DGID		|
						   IB_SA_PATH_REC_SGID		|
						   IB_SA_PATH_REC_NUMB_PATH	|
						   IB_SA_PATH_REC_PKEY,
						   SRP_PATH_REC_TIMEOUT_MS,
						   GFP_KERNEL,
						   srp_path_rec_completion,
						   target, &target->path_query);
	if (target->path_query_id < 0)
		return target->path_query_id;

	wait_for_completion(&target->done);

	if (target->status < 0)
338 339
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Path record query failed\n");
340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380

	return target->status;
}

static int srp_send_req(struct srp_target_port *target)
{
	struct {
		struct ib_cm_req_param param;
		struct srp_login_req   priv;
	} *req = NULL;
	int status;

	req = kzalloc(sizeof *req, GFP_KERNEL);
	if (!req)
		return -ENOMEM;

	req->param.primary_path 	      = &target->path;
	req->param.alternate_path 	      = NULL;
	req->param.service_id 		      = target->service_id;
	req->param.qp_num 		      = target->qp->qp_num;
	req->param.qp_type 		      = target->qp->qp_type;
	req->param.private_data 	      = &req->priv;
	req->param.private_data_len 	      = sizeof req->priv;
	req->param.flow_control 	      = 1;

	get_random_bytes(&req->param.starting_psn, 4);
	req->param.starting_psn 	     &= 0xffffff;

	/*
	 * Pick some arbitrary defaults here; we could make these
	 * module parameters if anyone cared about setting them.
	 */
	req->param.responder_resources	      = 4;
	req->param.remote_cm_response_timeout = 20;
	req->param.local_cm_response_timeout  = 20;
	req->param.retry_count 		      = 7;
	req->param.rnr_retry_count 	      = 7;
	req->param.max_cm_retries 	      = 15;

	req->priv.opcode     	= SRP_LOGIN_REQ;
	req->priv.tag        	= 0;
381
	req->priv.req_it_iu_len = cpu_to_be32(srp_max_iu_len);
382 383
	req->priv.req_buf_fmt 	= cpu_to_be16(SRP_BUF_FORMAT_DIRECT |
					      SRP_BUF_FORMAT_INDIRECT);
384
	/*
R
Roland Dreier 已提交
385
	 * In the published SRP specification (draft rev. 16a), the
386 387 388 389 390 391 392 393 394
	 * port identifier format is 8 bytes of ID extension followed
	 * by 8 bytes of GUID.  Older drafts put the two halves in the
	 * opposite order, so that the GUID comes first.
	 *
	 * Targets conforming to these obsolete drafts can be
	 * recognized by the I/O Class they report.
	 */
	if (target->io_class == SRP_REV10_IB_IO_CLASS) {
		memcpy(req->priv.initiator_port_id,
395
		       &target->path.sgid.global.interface_id, 8);
396
		memcpy(req->priv.initiator_port_id + 8,
397
		       &target->initiator_ext, 8);
398 399 400 401
		memcpy(req->priv.target_port_id,     &target->ioc_guid, 8);
		memcpy(req->priv.target_port_id + 8, &target->id_ext, 8);
	} else {
		memcpy(req->priv.initiator_port_id,
402 403 404
		       &target->initiator_ext, 8);
		memcpy(req->priv.initiator_port_id + 8,
		       &target->path.sgid.global.interface_id, 8);
405 406 407 408
		memcpy(req->priv.target_port_id,     &target->id_ext, 8);
		memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8);
	}

409 410
	/*
	 * Topspin/Cisco SRP targets will reject our login unless we
411 412
	 * zero out the first 8 bytes of our initiator port ID and set
	 * the second 8 bytes to the local node GUID.
413
	 */
414
	if (srp_target_is_topspin(target)) {
415 416 417 418
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Topspin/Cisco initiator port ID workaround "
			     "activated for target GUID %016llx\n",
			     (unsigned long long) be64_to_cpu(target->ioc_guid));
419
		memset(req->priv.initiator_port_id, 0, 8);
420
		memcpy(req->priv.initiator_port_id + 8,
421
		       &target->srp_host->srp_dev->dev->node_guid, 8);
422 423 424 425 426 427 428 429 430 431 432 433 434 435
	}

	status = ib_send_cm_req(target->cm_id, &req->param);

	kfree(req);

	return status;
}

static void srp_disconnect_target(struct srp_target_port *target)
{
	/* XXX should send SRP_I_LOGOUT request */

	init_completion(&target->done);
436
	if (ib_send_cm_dreq(target->cm_id, NULL, 0)) {
437 438
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM DREQ failed\n");
439 440
		return;
	}
441 442 443
	wait_for_completion(&target->done);
}

444 445 446 447 448 449
static bool srp_change_state(struct srp_target_port *target,
			    enum srp_target_state old,
			    enum srp_target_state new)
{
	bool changed = false;

450
	spin_lock_irq(&target->lock);
451 452 453 454
	if (target->state == old) {
		target->state = new;
		changed = true;
	}
455
	spin_unlock_irq(&target->lock);
456 457 458
	return changed;
}

D
David Howells 已提交
459
static void srp_remove_work(struct work_struct *work)
460
{
D
David Howells 已提交
461 462
	struct srp_target_port *target =
		container_of(work, struct srp_target_port, work);
463

464
	if (!srp_change_state(target, SRP_TARGET_DEAD, SRP_TARGET_REMOVED))
465 466
		return;

467
	spin_lock(&target->srp_host->target_lock);
468
	list_del(&target->list);
469
	spin_unlock(&target->srp_host->target_lock);
470

471
	srp_remove_host(target->scsi_host);
472 473 474 475 476 477 478 479
	scsi_remove_host(target->scsi_host);
	ib_destroy_cm_id(target->cm_id);
	srp_free_target_ib(target);
	scsi_host_put(target->scsi_host);
}

static int srp_connect_target(struct srp_target_port *target)
{
D
David Dillow 已提交
480
	int retries = 3;
481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512
	int ret;

	ret = srp_lookup_path(target);
	if (ret)
		return ret;

	while (1) {
		init_completion(&target->done);
		ret = srp_send_req(target);
		if (ret)
			return ret;
		wait_for_completion(&target->done);

		/*
		 * The CM event handling code will set status to
		 * SRP_PORT_REDIRECT if we get a port redirect REJ
		 * back, or SRP_DLID_REDIRECT if we get a lid/qp
		 * redirect REJ back.
		 */
		switch (target->status) {
		case 0:
			return 0;

		case SRP_PORT_REDIRECT:
			ret = srp_lookup_path(target);
			if (ret)
				return ret;
			break;

		case SRP_DLID_REDIRECT:
			break;

D
David Dillow 已提交
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527
		case SRP_STALE_CONN:
			/* Our current CM id was stale, and is now in timewait.
			 * Try to reconnect with a new one.
			 */
			if (!retries-- || srp_new_cm_id(target)) {
				shost_printk(KERN_ERR, target->scsi_host, PFX
					     "giving up on stale connection\n");
				target->status = -ECONNRESET;
				return target->status;
			}

			shost_printk(KERN_ERR, target->scsi_host, PFX
				     "retrying stale connection\n");
			break;

528 529 530 531 532 533
		default:
			return target->status;
		}
	}
}

534 535 536 537
static void srp_unmap_data(struct scsi_cmnd *scmnd,
			   struct srp_target_port *target,
			   struct srp_request *req)
{
538
	if (!scsi_sglist(scmnd) ||
539 540 541 542
	    (scmnd->sc_data_direction != DMA_TO_DEVICE &&
	     scmnd->sc_data_direction != DMA_FROM_DEVICE))
		return;

543 544 545 546 547
	if (req->fmr) {
		ib_fmr_pool_unmap(req->fmr);
		req->fmr = NULL;
	}

548
	ib_dma_unmap_sg(target->srp_host->srp_dev->dev, scsi_sglist(scmnd),
549
			scsi_sg_count(scmnd), scmnd->sc_data_direction);
550 551
}

552 553
static void srp_remove_req(struct srp_target_port *target,
			   struct srp_request *req, s32 req_lim_delta)
554
{
555 556
	unsigned long flags;

557
	srp_unmap_data(req->scmnd, target, req);
558
	spin_lock_irqsave(&target->lock, flags);
559
	target->req_lim += req_lim_delta;
560
	req->scmnd = NULL;
561
	list_add_tail(&req->list, &target->free_reqs);
562
	spin_unlock_irqrestore(&target->lock, flags);
563 564 565 566 567 568
}

static void srp_reset_req(struct srp_target_port *target, struct srp_request *req)
{
	req->scmnd->result = DID_RESET << 16;
	req->scmnd->scsi_done(req->scmnd);
569
	srp_remove_req(target, req, 0);
570 571
}

572 573 574 575
static int srp_reconnect_target(struct srp_target_port *target)
{
	struct ib_qp_attr qp_attr;
	struct ib_wc wc;
576
	int i, ret;
577

578
	if (!srp_change_state(target, SRP_TARGET_LIVE, SRP_TARGET_CONNECTING))
579 580 581 582 583 584 585
		return -EAGAIN;

	srp_disconnect_target(target);
	/*
	 * Now get a new local CM ID so that we avoid confusing the
	 * target in case things are really fouled up.
	 */
D
David Dillow 已提交
586 587
	ret = srp_new_cm_id(target);
	if (ret)
588 589 590 591 592 593 594 595 596 597 598
		goto err;

	qp_attr.qp_state = IB_QPS_RESET;
	ret = ib_modify_qp(target->qp, &qp_attr, IB_QP_STATE);
	if (ret)
		goto err;

	ret = srp_init_qp(target, target->qp);
	if (ret)
		goto err;

599 600 601
	while (ib_poll_cq(target->recv_cq, 1, &wc) > 0)
		; /* nothing */
	while (ib_poll_cq(target->send_cq, 1, &wc) > 0)
602 603
		; /* nothing */

604 605 606 607 608
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];
		if (req->scmnd)
			srp_reset_req(target, req);
	}
609

610
	INIT_LIST_HEAD(&target->free_tx);
611
	for (i = 0; i < SRP_SQ_SIZE; ++i)
612
		list_add(&target->tx_ring[i]->list, &target->free_tx);
613

614
	target->qp_in_error = 0;
615 616 617 618
	ret = srp_connect_target(target);
	if (ret)
		goto err;

619
	if (!srp_change_state(target, SRP_TARGET_CONNECTING, SRP_TARGET_LIVE))
620 621 622 623 624
		ret = -EAGAIN;

	return ret;

err:
625 626
	shost_printk(KERN_ERR, target->scsi_host,
		     PFX "reconnect failed (%d), removing target port.\n", ret);
627 628 629

	/*
	 * We couldn't reconnect, so kill our target port off.
630 631 632 633 634 635
	 * However, we have to defer the real removal because we
	 * are in the context of the SCSI error handler now, which
	 * will deadlock if we call scsi_remove_host().
	 *
	 * Schedule our work inside the lock to avoid a race with
	 * the flush_scheduled_work() in srp_remove_one().
636
	 */
637
	spin_lock_irq(&target->lock);
638 639
	if (target->state == SRP_TARGET_CONNECTING) {
		target->state = SRP_TARGET_DEAD;
D
David Howells 已提交
640
		INIT_WORK(&target->work, srp_remove_work);
641 642
		schedule_work(&target->work);
	}
643
	spin_unlock_irq(&target->lock);
644 645 646 647

	return ret;
}

648
static int srp_map_fmr(struct srp_target_port *target, struct scatterlist *scat,
649 650 651 652 653 654 655 656 657
		       int sg_cnt, struct srp_request *req,
		       struct srp_direct_buf *buf)
{
	u64 io_addr = 0;
	u64 *dma_pages;
	u32 len;
	int page_cnt;
	int i, j;
	int ret;
658
	struct srp_device *dev = target->srp_host->srp_dev;
659
	struct ib_device *ibdev = dev->dev;
660
	struct scatterlist *sg;
661 662 663 664

	if (!dev->fmr_pool)
		return -ENODEV;

665 666
	if (srp_target_is_mellanox(target) &&
	    (ib_sg_dma_address(ibdev, &scat[0]) & ~dev->fmr_page_mask))
667 668
		return -EINVAL;

669
	len = page_cnt = 0;
670 671
	scsi_for_each_sg(req->scmnd, sg, sg_cnt, i) {
		unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
672

673
		if (ib_sg_dma_address(ibdev, sg) & ~dev->fmr_page_mask) {
674 675 676 677 678
			if (i > 0)
				return -EINVAL;
			else
				++page_cnt;
		}
679
		if ((ib_sg_dma_address(ibdev, sg) + dma_len) &
680 681 682 683 684 685 686
		    ~dev->fmr_page_mask) {
			if (i < sg_cnt - 1)
				return -EINVAL;
			else
				++page_cnt;
		}

687
		len += dma_len;
688 689 690 691 692 693 694 695 696 697 698
	}

	page_cnt += len >> dev->fmr_page_shift;
	if (page_cnt > SRP_FMR_SIZE)
		return -ENOMEM;

	dma_pages = kmalloc(sizeof (u64) * page_cnt, GFP_ATOMIC);
	if (!dma_pages)
		return -ENOMEM;

	page_cnt = 0;
699 700
	scsi_for_each_sg(req->scmnd, sg, sg_cnt, i) {
		unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
701 702

		for (j = 0; j < dma_len; j += dev->fmr_page_size)
703
			dma_pages[page_cnt++] =
704
				(ib_sg_dma_address(ibdev, sg) &
705 706
				 dev->fmr_page_mask) + j;
	}
707 708

	req->fmr = ib_fmr_pool_map_phys(dev->fmr_pool,
709
					dma_pages, page_cnt, io_addr);
710 711
	if (IS_ERR(req->fmr)) {
		ret = PTR_ERR(req->fmr);
V
Vu Pham 已提交
712
		req->fmr = NULL;
713 714 715
		goto out;
	}

716 717
	buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, &scat[0]) &
			       ~dev->fmr_page_mask);
718 719 720 721 722 723 724 725 726 727 728
	buf->key = cpu_to_be32(req->fmr->fmr->rkey);
	buf->len = cpu_to_be32(len);

	ret = 0;

out:
	kfree(dma_pages);

	return ret;
}

729 730 731
static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
			struct srp_request *req)
{
732
	struct scatterlist *scat;
733
	struct srp_cmd *cmd = req->cmd->buf;
734
	int len, nents, count;
735
	u8 fmt = SRP_DATA_DESC_DIRECT;
736 737
	struct srp_device *dev;
	struct ib_device *ibdev;
738

739
	if (!scsi_sglist(scmnd) || scmnd->sc_data_direction == DMA_NONE)
740 741 742 743
		return sizeof (struct srp_cmd);

	if (scmnd->sc_data_direction != DMA_FROM_DEVICE &&
	    scmnd->sc_data_direction != DMA_TO_DEVICE) {
744 745 746
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled data direction %d\n",
			     scmnd->sc_data_direction);
747 748 749
		return -EINVAL;
	}

750 751
	nents = scsi_sg_count(scmnd);
	scat  = scsi_sglist(scmnd);
752

753
	dev = target->srp_host->srp_dev;
754 755 756
	ibdev = dev->dev;

	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
757 758 759

	fmt = SRP_DATA_DESC_DIRECT;
	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
760

761
	if (count == 1) {
762 763 764 765 766 767
		/*
		 * The midlayer only generated a single gather/scatter
		 * entry, or DMA mapping coalesced everything to a
		 * single entry.  So a direct descriptor along with
		 * the DMA MR suffices.
		 */
768
		struct srp_direct_buf *buf = (void *) cmd->add_data;
769

770
		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
771
		buf->key = cpu_to_be32(target->rkey);
772
		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
773
	} else if (srp_map_fmr(target, scat, count, req,
774 775 776 777 778 779
			       (void *) cmd->add_data)) {
		/*
		 * FMR mapping failed, and the scatterlist has more
		 * than one entry.  Generate an indirect memory
		 * descriptor.
		 */
780
		struct srp_indirect_buf *buf = (void *) cmd->add_data;
781
		struct scatterlist *sg;
782
		u32 datalen = 0;
783
		int i;
784

785
		fmt = SRP_DATA_DESC_INDIRECT;
786 787 788 789
		len = sizeof (struct srp_cmd) +
			sizeof (struct srp_indirect_buf) +
			count * sizeof (struct srp_direct_buf);

790 791
		scsi_for_each_sg(scmnd, sg, count, i) {
			unsigned int dma_len = ib_sg_dma_len(ibdev, sg);
792

793
			buf->desc_list[i].va  =
794
				cpu_to_be64(ib_sg_dma_address(ibdev, sg));
795
			buf->desc_list[i].key =
796
				cpu_to_be32(target->rkey);
797 798
			buf->desc_list[i].len = cpu_to_be32(dma_len);
			datalen += dma_len;
799
		}
800

801 802 803 804 805
		if (scmnd->sc_data_direction == DMA_TO_DEVICE)
			cmd->data_out_desc_cnt = count;
		else
			cmd->data_in_desc_cnt = count;

806 807
		buf->table_desc.va  =
			cpu_to_be64(req->cmd->dma + sizeof *cmd + sizeof *buf);
808
		buf->table_desc.key =
809
			cpu_to_be32(target->rkey);
810 811 812 813
		buf->table_desc.len =
			cpu_to_be32(count * sizeof (struct srp_direct_buf));

		buf->len = cpu_to_be32(datalen);
814 815 816 817 818 819 820 821 822 823
	}

	if (scmnd->sc_data_direction == DMA_TO_DEVICE)
		cmd->buf_fmt = fmt << 4;
	else
		cmd->buf_fmt = fmt;

	return len;
}

824 825 826 827 828 829 830 831
/*
 * Return an IU and possible credit to the free pool
 */
static void srp_put_tx_iu(struct srp_target_port *target, struct srp_iu *iu,
			  enum srp_iu_type iu_type)
{
	unsigned long flags;

832
	spin_lock_irqsave(&target->lock, flags);
833 834 835
	list_add(&iu->list, &target->free_tx);
	if (iu_type != SRP_IU_RSP)
		++target->req_lim;
836
	spin_unlock_irqrestore(&target->lock, flags);
837 838
}

839
/*
840 841
 * Must be called with target->lock held to protect req_lim and free_tx.
 * If IU is not sent, it must be returned using srp_put_tx_iu().
842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859
 *
 * Note:
 * An upper limit for the number of allocated information units for each
 * request type is:
 * - SRP_IU_CMD: SRP_CMD_SQ_SIZE, since the SCSI mid-layer never queues
 *   more than Scsi_Host.can_queue requests.
 * - SRP_IU_TSK_MGMT: SRP_TSK_MGMT_SQ_SIZE.
 * - SRP_IU_RSP: 1, since a conforming SRP target never sends more than
 *   one unanswered SRP request to an initiator.
 */
static struct srp_iu *__srp_get_tx_iu(struct srp_target_port *target,
				      enum srp_iu_type iu_type)
{
	s32 rsv = (iu_type == SRP_IU_TSK_MGMT) ? 0 : SRP_TSK_MGMT_SQ_SIZE;
	struct srp_iu *iu;

	srp_send_completion(target->send_cq, target);

860
	if (list_empty(&target->free_tx))
861 862 863
		return NULL;

	/* Initiator responses to target requests do not consume credits */
864 865 866 867 868 869 870
	if (iu_type != SRP_IU_RSP) {
		if (target->req_lim <= rsv) {
			++target->zero_req_lim;
			return NULL;
		}

		--target->req_lim;
871 872
	}

873
	iu = list_first_entry(&target->free_tx, struct srp_iu, list);
874
	list_del(&iu->list);
875 876 877
	return iu;
}

878 879
static int srp_post_send(struct srp_target_port *target,
			 struct srp_iu *iu, int len)
880 881 882 883 884 885
{
	struct ib_sge list;
	struct ib_send_wr wr, *bad_wr;

	list.addr   = iu->dma;
	list.length = len;
886
	list.lkey   = target->lkey;
887 888

	wr.next       = NULL;
889
	wr.wr_id      = (uintptr_t) iu;
890 891 892 893 894
	wr.sg_list    = &list;
	wr.num_sge    = 1;
	wr.opcode     = IB_WR_SEND;
	wr.send_flags = IB_SEND_SIGNALED;

895
	return ib_post_send(target->qp, &wr, &bad_wr);
896 897
}

898
static int srp_post_recv(struct srp_target_port *target, struct srp_iu *iu)
899 900
{
	struct ib_recv_wr wr, *bad_wr;
901
	struct ib_sge list;
902 903 904

	list.addr   = iu->dma;
	list.length = iu->size;
905
	list.lkey   = target->lkey;
906 907

	wr.next     = NULL;
908
	wr.wr_id    = (uintptr_t) iu;
909 910 911
	wr.sg_list  = &list;
	wr.num_sge  = 1;

912
	return ib_post_recv(target->qp, &wr, &bad_wr);
913 914
}

915 916 917 918 919 920 921
static void srp_process_rsp(struct srp_target_port *target, struct srp_rsp *rsp)
{
	struct srp_request *req;
	struct scsi_cmnd *scmnd;
	unsigned long flags;

	if (unlikely(rsp->tag & SRP_TAG_TSK_MGMT)) {
922
		spin_lock_irqsave(&target->lock, flags);
923
		target->req_lim += be32_to_cpu(rsp->req_lim_delta);
924
		spin_unlock_irqrestore(&target->lock, flags);
925

926 927 928 929
		target->tsk_mgmt_status = -1;
		if (be32_to_cpu(rsp->resp_data_len) >= 4)
			target->tsk_mgmt_status = rsp->data[3];
		complete(&target->tsk_mgmt_done);
930
	} else {
931
		req = &target->req_ring[rsp->tag];
932
		scmnd = req->scmnd;
933
		if (!scmnd)
934 935 936
			shost_printk(KERN_ERR, target->scsi_host,
				     "Null scmnd for RSP w/tag %016llx\n",
				     (unsigned long long) rsp->tag);
937 938 939 940 941 942 943 944 945 946
		scmnd->result = rsp->status;

		if (rsp->flags & SRP_RSP_FLAG_SNSVALID) {
			memcpy(scmnd->sense_buffer, rsp->data +
			       be32_to_cpu(rsp->resp_data_len),
			       min_t(int, be32_to_cpu(rsp->sense_data_len),
				     SCSI_SENSE_BUFFERSIZE));
		}

		if (rsp->flags & (SRP_RSP_FLAG_DOOVER | SRP_RSP_FLAG_DOUNDER))
947
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_out_res_cnt));
948
		else if (rsp->flags & (SRP_RSP_FLAG_DIOVER | SRP_RSP_FLAG_DIUNDER))
949
			scsi_set_resid(scmnd, be32_to_cpu(rsp->data_in_res_cnt));
950

951
		srp_remove_req(target, req, be32_to_cpu(rsp->req_lim_delta));
952 953
		scmnd->host_scribble = NULL;
		scmnd->scsi_done(scmnd);
954 955 956
	}
}

957 958 959
static int srp_response_common(struct srp_target_port *target, s32 req_delta,
			       void *rsp, int len)
{
960
	struct ib_device *dev = target->srp_host->srp_dev->dev;
961 962
	unsigned long flags;
	struct srp_iu *iu;
963
	int err;
964

965
	spin_lock_irqsave(&target->lock, flags);
966 967
	target->req_lim += req_delta;
	iu = __srp_get_tx_iu(target, SRP_IU_RSP);
968
	spin_unlock_irqrestore(&target->lock, flags);
969

970 971 972
	if (!iu) {
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "no IU available to send response\n");
973
		return 1;
974 975 976 977 978 979
	}

	ib_dma_sync_single_for_cpu(dev, iu->dma, len, DMA_TO_DEVICE);
	memcpy(iu->buf, rsp, len);
	ib_dma_sync_single_for_device(dev, iu->dma, len, DMA_TO_DEVICE);

980 981
	err = srp_post_send(target, iu, len);
	if (err) {
982 983
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "unable to post response: %d\n", err);
984 985
		srp_put_tx_iu(target, iu, SRP_IU_RSP);
	}
986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020

	return err;
}

static void srp_process_cred_req(struct srp_target_port *target,
				 struct srp_cred_req *req)
{
	struct srp_cred_rsp rsp = {
		.opcode = SRP_CRED_RSP,
		.tag = req->tag,
	};
	s32 delta = be32_to_cpu(req->req_lim_delta);

	if (srp_response_common(target, delta, &rsp, sizeof rsp))
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "problems processing SRP_CRED_REQ\n");
}

static void srp_process_aer_req(struct srp_target_port *target,
				struct srp_aer_req *req)
{
	struct srp_aer_rsp rsp = {
		.opcode = SRP_AER_RSP,
		.tag = req->tag,
	};
	s32 delta = be32_to_cpu(req->req_lim_delta);

	shost_printk(KERN_ERR, target->scsi_host, PFX
		     "ignoring AER for LUN %llu\n", be64_to_cpu(req->lun));

	if (srp_response_common(target, delta, &rsp, sizeof rsp))
		shost_printk(KERN_ERR, target->scsi_host, PFX
			     "problems processing SRP_AER_REQ\n");
}

1021 1022
static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
{
1023 1024
	struct ib_device *dev = target->srp_host->srp_dev->dev;
	struct srp_iu *iu = (struct srp_iu *) wc->wr_id;
1025
	int res;
1026 1027
	u8 opcode;

1028 1029
	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_ti_iu_len,
				   DMA_FROM_DEVICE);
1030 1031 1032 1033

	opcode = *(u8 *) iu->buf;

	if (0) {
1034 1035
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "recv completion, opcode 0x%02x\n", opcode);
B
Bart Van Assche 已提交
1036 1037
		print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 8, 1,
			       iu->buf, wc->byte_len, true);
1038 1039 1040 1041 1042 1043 1044
	}

	switch (opcode) {
	case SRP_RSP:
		srp_process_rsp(target, iu->buf);
		break;

1045 1046 1047 1048 1049 1050 1051 1052
	case SRP_CRED_REQ:
		srp_process_cred_req(target, iu->buf);
		break;

	case SRP_AER_REQ:
		srp_process_aer_req(target, iu->buf);
		break;

1053 1054
	case SRP_T_LOGOUT:
		/* XXX Handle target logout */
1055 1056
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Got target logout request\n");
1057 1058 1059
		break;

	default:
1060 1061
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled SRP opcode 0x%02x\n", opcode);
1062 1063 1064
		break;
	}

1065 1066
	ib_dma_sync_single_for_device(dev, iu->dma, target->max_ti_iu_len,
				      DMA_FROM_DEVICE);
1067

1068
	res = srp_post_recv(target, iu);
1069 1070 1071
	if (res != 0)
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Recv failed with error code %d\n", res);
1072 1073
}

1074
static void srp_recv_completion(struct ib_cq *cq, void *target_ptr)
1075 1076 1077 1078 1079 1080 1081
{
	struct srp_target_port *target = target_ptr;
	struct ib_wc wc;

	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
	while (ib_poll_cq(cq, 1, &wc) > 0) {
		if (wc.status) {
1082
			shost_printk(KERN_ERR, target->scsi_host,
1083
				     PFX "failed receive status %d\n",
1084
				     wc.status);
1085
			target->qp_in_error = 1;
1086 1087 1088
			break;
		}

1089 1090 1091 1092 1093 1094 1095 1096
		srp_handle_recv(target, &wc);
	}
}

static void srp_send_completion(struct ib_cq *cq, void *target_ptr)
{
	struct srp_target_port *target = target_ptr;
	struct ib_wc wc;
1097
	struct srp_iu *iu;
1098 1099 1100 1101 1102 1103 1104 1105 1106 1107

	while (ib_poll_cq(cq, 1, &wc) > 0) {
		if (wc.status) {
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "failed send status %d\n",
				     wc.status);
			target->qp_in_error = 1;
			break;
		}

1108 1109
		iu = (struct srp_iu *) wc.wr_id;
		list_add(&iu->list, &target->free_tx);
1110 1111 1112
	}
}

1113
static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd)
1114
{
1115
	struct srp_target_port *target = host_to_target(shost);
1116 1117 1118
	struct srp_request *req;
	struct srp_iu *iu;
	struct srp_cmd *cmd;
1119
	struct ib_device *dev;
1120
	unsigned long flags;
1121 1122 1123 1124 1125 1126 1127 1128
	int len;

	if (target->state == SRP_TARGET_CONNECTING)
		goto err;

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED) {
		scmnd->result = DID_BAD_TARGET << 16;
1129
		scmnd->scsi_done(scmnd);
1130 1131 1132
		return 0;
	}

1133
	spin_lock_irqsave(&target->lock, flags);
1134
	iu = __srp_get_tx_iu(target, SRP_IU_CMD);
1135 1136 1137 1138 1139
	if (iu) {
		req = list_first_entry(&target->free_reqs, struct srp_request,
				      list);
		list_del(&req->list);
	}
1140
	spin_unlock_irqrestore(&target->lock, flags);
1141

1142 1143 1144
	if (!iu)
		goto err;

1145
	dev = target->srp_host->srp_dev->dev;
1146 1147
	ib_dma_sync_single_for_cpu(dev, iu->dma, srp_max_iu_len,
				   DMA_TO_DEVICE);
1148 1149

	scmnd->result        = 0;
1150
	scmnd->host_scribble = (void *) req;
1151 1152 1153 1154 1155 1156

	cmd = iu->buf;
	memset(cmd, 0, sizeof *cmd);

	cmd->opcode = SRP_CMD;
	cmd->lun    = cpu_to_be64((u64) scmnd->device->lun << 48);
1157
	cmd->tag    = req->index;
1158 1159 1160 1161 1162 1163 1164
	memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len);

	req->scmnd    = scmnd;
	req->cmd      = iu;

	len = srp_map_data(scmnd, target, req);
	if (len < 0) {
1165 1166
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Failed to map data\n");
1167
		goto err_iu;
1168 1169
	}

1170 1171
	ib_dma_sync_single_for_device(dev, iu->dma, srp_max_iu_len,
				      DMA_TO_DEVICE);
1172

1173
	if (srp_post_send(target, iu, len)) {
1174
		shost_printk(KERN_ERR, target->scsi_host, PFX "Send failed\n");
1175 1176 1177 1178 1179 1180 1181 1182
		goto err_unmap;
	}

	return 0;

err_unmap:
	srp_unmap_data(scmnd, target, req);

1183 1184 1185
err_iu:
	srp_put_tx_iu(target, iu, SRP_IU_CMD);

1186
	spin_lock_irqsave(&target->lock, flags);
1187
	list_add(&req->list, &target->free_reqs);
1188
	spin_unlock_irqrestore(&target->lock, flags);
1189

1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205
err:
	return SCSI_MLQUEUE_HOST_BUSY;
}

static int srp_alloc_iu_bufs(struct srp_target_port *target)
{
	int i;

	for (i = 0; i < SRP_RQ_SIZE; ++i) {
		target->rx_ring[i] = srp_alloc_iu(target->srp_host,
						  target->max_ti_iu_len,
						  GFP_KERNEL, DMA_FROM_DEVICE);
		if (!target->rx_ring[i])
			goto err;
	}

1206
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1207
		target->tx_ring[i] = srp_alloc_iu(target->srp_host,
1208
						  srp_max_iu_len,
1209 1210 1211
						  GFP_KERNEL, DMA_TO_DEVICE);
		if (!target->tx_ring[i])
			goto err;
1212 1213

		list_add(&target->tx_ring[i]->list, &target->free_tx);
1214 1215 1216 1217 1218 1219 1220 1221 1222 1223
	}

	return 0;

err:
	for (i = 0; i < SRP_RQ_SIZE; ++i) {
		srp_free_iu(target->srp_host, target->rx_ring[i]);
		target->rx_ring[i] = NULL;
	}

1224
	for (i = 0; i < SRP_SQ_SIZE; ++i) {
1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235
		srp_free_iu(target->srp_host, target->tx_ring[i]);
		target->tx_ring[i] = NULL;
	}

	return -ENOMEM;
}

static void srp_cm_rej_handler(struct ib_cm_id *cm_id,
			       struct ib_cm_event *event,
			       struct srp_target_port *target)
{
1236
	struct Scsi_Host *shost = target->scsi_host;
1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252
	struct ib_class_port_info *cpi;
	int opcode;

	switch (event->param.rej_rcvd.reason) {
	case IB_CM_REJ_PORT_CM_REDIRECT:
		cpi = event->param.rej_rcvd.ari;
		target->path.dlid = cpi->redirect_lid;
		target->path.pkey = cpi->redirect_pkey;
		cm_id->remote_cm_qpn = be32_to_cpu(cpi->redirect_qp) & 0x00ffffff;
		memcpy(target->path.dgid.raw, cpi->redirect_gid, 16);

		target->status = target->path.dlid ?
			SRP_DLID_REDIRECT : SRP_PORT_REDIRECT;
		break;

	case IB_CM_REJ_PORT_REDIRECT:
1253
		if (srp_target_is_topspin(target)) {
1254 1255 1256 1257 1258 1259 1260 1261
			/*
			 * Topspin/Cisco SRP gateways incorrectly send
			 * reject reason code 25 when they mean 24
			 * (port redirect).
			 */
			memcpy(target->path.dgid.raw,
			       event->param.rej_rcvd.ari, 16);

1262 1263 1264 1265
			shost_printk(KERN_DEBUG, shost,
				     PFX "Topspin/Cisco redirect to target port GID %016llx%016llx\n",
				     (unsigned long long) be64_to_cpu(target->path.dgid.global.subnet_prefix),
				     (unsigned long long) be64_to_cpu(target->path.dgid.global.interface_id));
1266 1267 1268

			target->status = SRP_PORT_REDIRECT;
		} else {
1269 1270
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_PORT_REDIRECT\n");
1271 1272 1273 1274 1275
			target->status = -ECONNRESET;
		}
		break;

	case IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID:
1276 1277
		shost_printk(KERN_WARNING, shost,
			    "  REJ reason: IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID\n");
1278 1279 1280 1281 1282 1283 1284 1285 1286 1287
		target->status = -ECONNRESET;
		break;

	case IB_CM_REJ_CONSUMER_DEFINED:
		opcode = *(u8 *) event->private_data;
		if (opcode == SRP_LOGIN_REJ) {
			struct srp_login_rej *rej = event->private_data;
			u32 reason = be32_to_cpu(rej->reason);

			if (reason == SRP_LOGIN_REJ_REQ_IT_IU_LENGTH_TOO_LARGE)
1288 1289
				shost_printk(KERN_WARNING, shost,
					     PFX "SRP_LOGIN_REJ: requested max_it_iu_len too large\n");
1290
			else
1291 1292
				shost_printk(KERN_WARNING, shost,
					    PFX "SRP LOGIN REJECTED, reason 0x%08x\n", reason);
1293
		} else
1294 1295 1296
			shost_printk(KERN_WARNING, shost,
				     "  REJ reason: IB_CM_REJ_CONSUMER_DEFINED,"
				     " opcode 0x%02x\n", opcode);
1297 1298 1299
		target->status = -ECONNRESET;
		break;

D
David Dillow 已提交
1300 1301 1302 1303 1304
	case IB_CM_REJ_STALE_CONN:
		shost_printk(KERN_WARNING, shost, "  REJ reason: stale connection\n");
		target->status = SRP_STALE_CONN;
		break;

1305
	default:
1306 1307
		shost_printk(KERN_WARNING, shost, "  REJ reason 0x%x\n",
			     event->param.rej_rcvd.reason);
1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
		target->status = -ECONNRESET;
	}
}

static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event)
{
	struct srp_target_port *target = cm_id->context;
	struct ib_qp_attr *qp_attr = NULL;
	int attr_mask = 0;
	int comp = 0;
	int opcode = 0;
1319
	int i;
1320 1321 1322

	switch (event->event) {
	case IB_CM_REQ_ERROR:
1323 1324
		shost_printk(KERN_DEBUG, target->scsi_host,
			     PFX "Sending CM REQ failed\n");
1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338
		comp = 1;
		target->status = -ECONNRESET;
		break;

	case IB_CM_REP_RECEIVED:
		comp = 1;
		opcode = *(u8 *) event->private_data;

		if (opcode == SRP_LOGIN_RSP) {
			struct srp_login_rsp *rsp = event->private_data;

			target->max_ti_iu_len = be32_to_cpu(rsp->max_ti_iu_len);
			target->req_lim       = be32_to_cpu(rsp->req_lim_delta);

1339 1340 1341 1342 1343 1344 1345
			/*
			 * Reserve credits for task management so we don't
			 * bounce requests back to the SCSI mid-layer.
			 */
			target->scsi_host->can_queue
				= min(target->req_lim - SRP_TSK_MGMT_SQ_SIZE,
				      target->scsi_host->can_queue);
1346
		} else {
1347 1348
			shost_printk(KERN_WARNING, target->scsi_host,
				    PFX "Unhandled RSP opcode %#x\n", opcode);
1349 1350 1351 1352
			target->status = -ECONNRESET;
			break;
		}

V
Vu Pham 已提交
1353 1354 1355 1356 1357
		if (!target->rx_ring[0]) {
			target->status = srp_alloc_iu_bufs(target);
			if (target->status)
				break;
		}
1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373

		qp_attr = kmalloc(sizeof *qp_attr, GFP_KERNEL);
		if (!qp_attr) {
			target->status = -ENOMEM;
			break;
		}

		qp_attr->qp_state = IB_QPS_RTR;
		target->status = ib_cm_init_qp_attr(cm_id, qp_attr, &attr_mask);
		if (target->status)
			break;

		target->status = ib_modify_qp(target->qp, qp_attr, attr_mask);
		if (target->status)
			break;

1374
		for (i = 0; i < SRP_RQ_SIZE; i++) {
1375 1376
			struct srp_iu *iu = target->rx_ring[i];
			target->status = srp_post_recv(target, iu);
1377 1378 1379
			if (target->status)
				break;
		}
1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398
		if (target->status)
			break;

		qp_attr->qp_state = IB_QPS_RTS;
		target->status = ib_cm_init_qp_attr(cm_id, qp_attr, &attr_mask);
		if (target->status)
			break;

		target->status = ib_modify_qp(target->qp, qp_attr, attr_mask);
		if (target->status)
			break;

		target->status = ib_send_cm_rtu(cm_id, NULL, 0);
		if (target->status)
			break;

		break;

	case IB_CM_REJ_RECEIVED:
1399
		shost_printk(KERN_DEBUG, target->scsi_host, PFX "REJ received\n");
1400 1401 1402 1403 1404
		comp = 1;

		srp_cm_rej_handler(cm_id, event, target);
		break;

1405
	case IB_CM_DREQ_RECEIVED:
1406 1407
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "DREQ received - connection closed\n");
1408
		if (ib_send_cm_drep(cm_id, NULL, 0))
1409 1410
			shost_printk(KERN_ERR, target->scsi_host,
				     PFX "Sending CM DREP failed\n");
1411 1412 1413
		break;

	case IB_CM_TIMEWAIT_EXIT:
1414 1415
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "connection closed\n");
1416 1417 1418 1419 1420

		comp = 1;
		target->status = 0;
		break;

1421 1422 1423 1424 1425
	case IB_CM_MRA_RECEIVED:
	case IB_CM_DREQ_ERROR:
	case IB_CM_DREP_RECEIVED:
		break;

1426
	default:
1427 1428
		shost_printk(KERN_WARNING, target->scsi_host,
			     PFX "Unhandled CM event %d\n", event->event);
1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439
		break;
	}

	if (comp)
		complete(&target->done);

	kfree(qp_attr);

	return 0;
}

1440
static int srp_send_tsk_mgmt(struct srp_target_port *target,
1441
			     u64 req_tag, unsigned int lun, u8 func)
1442
{
1443
	struct ib_device *dev = target->srp_host->srp_dev->dev;
1444 1445 1446
	struct srp_iu *iu;
	struct srp_tsk_mgmt *tsk_mgmt;

1447
	if (target->state == SRP_TARGET_DEAD ||
1448
	    target->state == SRP_TARGET_REMOVED)
1449
		return -1;
1450

1451
	init_completion(&target->tsk_mgmt_done);
1452

1453
	spin_lock_irq(&target->lock);
1454
	iu = __srp_get_tx_iu(target, SRP_IU_TSK_MGMT);
1455
	spin_unlock_irq(&target->lock);
1456

1457
	if (!iu)
1458
		return -1;
1459

1460 1461
	ib_dma_sync_single_for_cpu(dev, iu->dma, sizeof *tsk_mgmt,
				   DMA_TO_DEVICE);
1462 1463 1464 1465
	tsk_mgmt = iu->buf;
	memset(tsk_mgmt, 0, sizeof *tsk_mgmt);

	tsk_mgmt->opcode 	= SRP_TSK_MGMT;
1466 1467
	tsk_mgmt->lun		= cpu_to_be64((u64) lun << 48);
	tsk_mgmt->tag		= req_tag | SRP_TAG_TSK_MGMT;
1468
	tsk_mgmt->tsk_mgmt_func = func;
1469
	tsk_mgmt->task_tag	= req_tag;
1470

1471 1472
	ib_dma_sync_single_for_device(dev, iu->dma, sizeof *tsk_mgmt,
				      DMA_TO_DEVICE);
1473 1474 1475 1476
	if (srp_post_send(target, iu, sizeof *tsk_mgmt)) {
		srp_put_tx_iu(target, iu, SRP_IU_TSK_MGMT);
		return -1;
	}
1477

1478
	if (!wait_for_completion_timeout(&target->tsk_mgmt_done,
1479
					 msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS)))
1480
		return -1;
1481

1482 1483 1484
	return 0;
}

1485 1486
static int srp_abort(struct scsi_cmnd *scmnd)
{
1487
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1488
	struct srp_request *req = (struct srp_request *) scmnd->host_scribble;
1489 1490
	int ret = SUCCESS;

1491
	shost_printk(KERN_ERR, target->scsi_host, "SRP abort called\n");
1492

1493
	if (!req || target->qp_in_error)
1494
		return FAILED;
1495 1496
	if (srp_send_tsk_mgmt(target, req->index, scmnd->device->lun,
			      SRP_TSK_ABORT_TASK))
1497 1498
		return FAILED;

1499 1500
	if (req->scmnd) {
		if (!target->tsk_mgmt_status) {
1501
			srp_remove_req(target, req, 0);
1502 1503 1504 1505
			scmnd->result = DID_ABORT << 16;
		} else
			ret = FAILED;
	}
1506 1507

	return ret;
1508 1509 1510 1511
}

static int srp_reset_device(struct scsi_cmnd *scmnd)
{
1512
	struct srp_target_port *target = host_to_target(scmnd->device->host);
1513
	int i;
1514

1515
	shost_printk(KERN_ERR, target->scsi_host, "SRP reset_device called\n");
1516

1517 1518
	if (target->qp_in_error)
		return FAILED;
1519 1520
	if (srp_send_tsk_mgmt(target, SRP_TAG_NO_REQ, scmnd->device->lun,
			      SRP_TSK_LUN_RESET))
1521
		return FAILED;
1522
	if (target->tsk_mgmt_status)
1523 1524
		return FAILED;

1525 1526
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
		struct srp_request *req = &target->req_ring[i];
1527
		if (req->scmnd && req->scmnd->device == scmnd->device)
1528
			srp_reset_req(target, req);
1529
	}
1530 1531

	return SUCCESS;
1532 1533 1534 1535 1536 1537 1538
}

static int srp_reset_host(struct scsi_cmnd *scmnd)
{
	struct srp_target_port *target = host_to_target(scmnd->device->host);
	int ret = FAILED;

1539
	shost_printk(KERN_ERR, target->scsi_host, PFX "SRP reset_host called\n");
1540 1541 1542 1543 1544 1545 1546

	if (!srp_reconnect_target(target))
		ret = SUCCESS;

	return ret;
}

1547 1548
static ssize_t show_id_ext(struct device *dev, struct device_attribute *attr,
			   char *buf)
1549
{
1550
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1551 1552 1553 1554 1555 1556 1557 1558 1559

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

	return sprintf(buf, "0x%016llx\n",
		       (unsigned long long) be64_to_cpu(target->id_ext));
}

1560 1561
static ssize_t show_ioc_guid(struct device *dev, struct device_attribute *attr,
			     char *buf)
1562
{
1563
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1564 1565 1566 1567 1568 1569 1570 1571 1572

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

	return sprintf(buf, "0x%016llx\n",
		       (unsigned long long) be64_to_cpu(target->ioc_guid));
}

1573 1574
static ssize_t show_service_id(struct device *dev,
			       struct device_attribute *attr, char *buf)
1575
{
1576
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1577 1578 1579 1580 1581 1582 1583 1584 1585

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

	return sprintf(buf, "0x%016llx\n",
		       (unsigned long long) be64_to_cpu(target->service_id));
}

1586 1587
static ssize_t show_pkey(struct device *dev, struct device_attribute *attr,
			 char *buf)
1588
{
1589
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1590 1591 1592 1593 1594 1595 1596 1597

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

	return sprintf(buf, "0x%04x\n", be16_to_cpu(target->path.pkey));
}

1598 1599
static ssize_t show_dgid(struct device *dev, struct device_attribute *attr,
			 char *buf)
1600
{
1601
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1602 1603 1604 1605 1606

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

H
Harvey Harrison 已提交
1607
	return sprintf(buf, "%pI6\n", target->path.dgid.raw);
1608 1609
}

1610 1611
static ssize_t show_orig_dgid(struct device *dev,
			      struct device_attribute *attr, char *buf)
1612
{
1613
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1614 1615 1616 1617 1618

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

H
Harvey Harrison 已提交
1619
	return sprintf(buf, "%pI6\n", target->orig_dgid);
1620 1621
}

1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633
static ssize_t show_req_lim(struct device *dev,
			    struct device_attribute *attr, char *buf)
{
	struct srp_target_port *target = host_to_target(class_to_shost(dev));

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

	return sprintf(buf, "%d\n", target->req_lim);
}

1634 1635
static ssize_t show_zero_req_lim(struct device *dev,
				 struct device_attribute *attr, char *buf)
1636
{
1637
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1638 1639 1640 1641 1642 1643 1644 1645

	if (target->state == SRP_TARGET_DEAD ||
	    target->state == SRP_TARGET_REMOVED)
		return -ENODEV;

	return sprintf(buf, "%d\n", target->zero_req_lim);
}

1646 1647
static ssize_t show_local_ib_port(struct device *dev,
				  struct device_attribute *attr, char *buf)
1648
{
1649
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1650 1651 1652 1653

	return sprintf(buf, "%d\n", target->srp_host->port);
}

1654 1655
static ssize_t show_local_ib_device(struct device *dev,
				    struct device_attribute *attr, char *buf)
1656
{
1657
	struct srp_target_port *target = host_to_target(class_to_shost(dev));
1658

1659
	return sprintf(buf, "%s\n", target->srp_host->srp_dev->dev->name);
1660 1661
}

1662 1663 1664 1665 1666 1667
static DEVICE_ATTR(id_ext,	    S_IRUGO, show_id_ext,	   NULL);
static DEVICE_ATTR(ioc_guid,	    S_IRUGO, show_ioc_guid,	   NULL);
static DEVICE_ATTR(service_id,	    S_IRUGO, show_service_id,	   NULL);
static DEVICE_ATTR(pkey,	    S_IRUGO, show_pkey,		   NULL);
static DEVICE_ATTR(dgid,	    S_IRUGO, show_dgid,		   NULL);
static DEVICE_ATTR(orig_dgid,	    S_IRUGO, show_orig_dgid,	   NULL);
1668
static DEVICE_ATTR(req_lim,         S_IRUGO, show_req_lim,         NULL);
1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679
static DEVICE_ATTR(zero_req_lim,    S_IRUGO, show_zero_req_lim,	   NULL);
static DEVICE_ATTR(local_ib_port,   S_IRUGO, show_local_ib_port,   NULL);
static DEVICE_ATTR(local_ib_device, S_IRUGO, show_local_ib_device, NULL);

static struct device_attribute *srp_host_attrs[] = {
	&dev_attr_id_ext,
	&dev_attr_ioc_guid,
	&dev_attr_service_id,
	&dev_attr_pkey,
	&dev_attr_dgid,
	&dev_attr_orig_dgid,
1680
	&dev_attr_req_lim,
1681 1682 1683
	&dev_attr_zero_req_lim,
	&dev_attr_local_ib_port,
	&dev_attr_local_ib_device,
1684 1685 1686
	NULL
};

1687 1688
static struct scsi_host_template srp_template = {
	.module				= THIS_MODULE,
R
Roland Dreier 已提交
1689 1690
	.name				= "InfiniBand SRP initiator",
	.proc_name			= DRV_NAME,
1691 1692 1693 1694 1695
	.info				= srp_target_info,
	.queuecommand			= srp_queuecommand,
	.eh_abort_handler		= srp_abort,
	.eh_device_reset_handler	= srp_reset_device,
	.eh_host_reset_handler		= srp_reset_host,
1696
	.can_queue			= SRP_CMD_SQ_SIZE,
1697
	.this_id			= -1,
1698
	.cmd_per_lun			= SRP_CMD_SQ_SIZE,
1699 1700
	.use_clustering			= ENABLE_CLUSTERING,
	.shost_attrs			= srp_host_attrs
1701 1702 1703 1704
};

static int srp_add_target(struct srp_host *host, struct srp_target_port *target)
{
1705 1706 1707
	struct srp_rport_identifiers ids;
	struct srp_rport *rport;

1708 1709 1710
	sprintf(target->target_name, "SRP.T10:%016llX",
		 (unsigned long long) be64_to_cpu(target->id_ext));

1711
	if (scsi_add_host(target->scsi_host, host->srp_dev->dev->dma_device))
1712 1713
		return -ENODEV;

1714 1715
	memcpy(ids.port_id, &target->id_ext, 8);
	memcpy(ids.port_id + 8, &target->ioc_guid, 8);
1716
	ids.roles = SRP_RPORT_ROLE_TARGET;
1717 1718 1719 1720 1721 1722
	rport = srp_rport_add(target->scsi_host, &ids);
	if (IS_ERR(rport)) {
		scsi_remove_host(target->scsi_host);
		return PTR_ERR(rport);
	}

1723
	spin_lock(&host->target_lock);
1724
	list_add_tail(&target->list, &host->target_list);
1725
	spin_unlock(&host->target_lock);
1726 1727 1728 1729

	target->state = SRP_TARGET_LIVE;

	scsi_scan_target(&target->scsi_host->shost_gendev,
1730
			 0, target->scsi_id, SCAN_WILD_CARD, 0);
1731 1732 1733 1734

	return 0;
}

1735
static void srp_release_dev(struct device *dev)
1736 1737
{
	struct srp_host *host =
1738
		container_of(dev, struct srp_host, dev);
1739 1740 1741 1742 1743 1744

	complete(&host->released);
}

static struct class srp_class = {
	.name    = "infiniband_srp",
1745
	.dev_release = srp_release_dev
1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763
};

/*
 * Target ports are added by writing
 *
 *     id_ext=<SRP ID ext>,ioc_guid=<SRP IOC GUID>,dgid=<dest GID>,
 *     pkey=<P_Key>,service_id=<service ID>
 *
 * to the add_target sysfs attribute.
 */
enum {
	SRP_OPT_ERR		= 0,
	SRP_OPT_ID_EXT		= 1 << 0,
	SRP_OPT_IOC_GUID	= 1 << 1,
	SRP_OPT_DGID		= 1 << 2,
	SRP_OPT_PKEY		= 1 << 3,
	SRP_OPT_SERVICE_ID	= 1 << 4,
	SRP_OPT_MAX_SECT	= 1 << 5,
1764
	SRP_OPT_MAX_CMD_PER_LUN	= 1 << 6,
1765
	SRP_OPT_IO_CLASS	= 1 << 7,
1766
	SRP_OPT_INITIATOR_EXT	= 1 << 8,
1767 1768 1769 1770 1771 1772 1773
	SRP_OPT_ALL		= (SRP_OPT_ID_EXT	|
				   SRP_OPT_IOC_GUID	|
				   SRP_OPT_DGID		|
				   SRP_OPT_PKEY		|
				   SRP_OPT_SERVICE_ID),
};

1774
static const match_table_t srp_opt_tokens = {
1775 1776 1777 1778 1779 1780 1781
	{ SRP_OPT_ID_EXT,		"id_ext=%s" 		},
	{ SRP_OPT_IOC_GUID,		"ioc_guid=%s" 		},
	{ SRP_OPT_DGID,			"dgid=%s" 		},
	{ SRP_OPT_PKEY,			"pkey=%x" 		},
	{ SRP_OPT_SERVICE_ID,		"service_id=%s"		},
	{ SRP_OPT_MAX_SECT,		"max_sect=%d" 		},
	{ SRP_OPT_MAX_CMD_PER_LUN,	"max_cmd_per_lun=%d" 	},
1782
	{ SRP_OPT_IO_CLASS,		"io_class=%x"		},
1783
	{ SRP_OPT_INITIATOR_EXT,	"initiator_ext=%s"	},
1784
	{ SRP_OPT_ERR,			NULL 			}
1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812
};

static int srp_parse_options(const char *buf, struct srp_target_port *target)
{
	char *options, *sep_opt;
	char *p;
	char dgid[3];
	substring_t args[MAX_OPT_ARGS];
	int opt_mask = 0;
	int token;
	int ret = -EINVAL;
	int i;

	options = kstrdup(buf, GFP_KERNEL);
	if (!options)
		return -ENOMEM;

	sep_opt = options;
	while ((p = strsep(&sep_opt, ",")) != NULL) {
		if (!*p)
			continue;

		token = match_token(p, srp_opt_tokens, args);
		opt_mask |= token;

		switch (token) {
		case SRP_OPT_ID_EXT:
			p = match_strdup(args);
1813 1814 1815 1816
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1817 1818 1819 1820 1821 1822
			target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_IOC_GUID:
			p = match_strdup(args);
1823 1824 1825 1826
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1827 1828 1829 1830 1831 1832
			target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

		case SRP_OPT_DGID:
			p = match_strdup(args);
1833 1834 1835 1836
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1837 1838
			if (strlen(p) != 32) {
				printk(KERN_WARNING PFX "bad dest GID parameter '%s'\n", p);
1839
				kfree(p);
1840 1841 1842 1843 1844 1845 1846
				goto out;
			}

			for (i = 0; i < 16; ++i) {
				strlcpy(dgid, p + i * 2, 3);
				target->path.dgid.raw[i] = simple_strtoul(dgid, NULL, 16);
			}
1847
			kfree(p);
1848
			memcpy(target->orig_dgid, target->path.dgid.raw, 16);
1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860
			break;

		case SRP_OPT_PKEY:
			if (match_hex(args, &token)) {
				printk(KERN_WARNING PFX "bad P_Key parameter '%s'\n", p);
				goto out;
			}
			target->path.pkey = cpu_to_be16(token);
			break;

		case SRP_OPT_SERVICE_ID:
			p = match_strdup(args);
1861 1862 1863 1864
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1865
			target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16));
1866
			target->path.service_id = target->service_id;
1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877
			kfree(p);
			break;

		case SRP_OPT_MAX_SECT:
			if (match_int(args, &token)) {
				printk(KERN_WARNING PFX "bad max sect parameter '%s'\n", p);
				goto out;
			}
			target->scsi_host->max_sectors = token;
			break;

1878 1879 1880 1881 1882
		case SRP_OPT_MAX_CMD_PER_LUN:
			if (match_int(args, &token)) {
				printk(KERN_WARNING PFX "bad max cmd_per_lun parameter '%s'\n", p);
				goto out;
			}
1883
			target->scsi_host->cmd_per_lun = min(token, SRP_CMD_SQ_SIZE);
1884 1885
			break;

1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900
		case SRP_OPT_IO_CLASS:
			if (match_hex(args, &token)) {
				printk(KERN_WARNING PFX "bad  IO class parameter '%s' \n", p);
				goto out;
			}
			if (token != SRP_REV10_IB_IO_CLASS &&
			    token != SRP_REV16A_IB_IO_CLASS) {
				printk(KERN_WARNING PFX "unknown IO class parameter value"
				       " %x specified (use %x or %x).\n",
				       token, SRP_REV10_IB_IO_CLASS, SRP_REV16A_IB_IO_CLASS);
				goto out;
			}
			target->io_class = token;
			break;

1901 1902
		case SRP_OPT_INITIATOR_EXT:
			p = match_strdup(args);
1903 1904 1905 1906
			if (!p) {
				ret = -ENOMEM;
				goto out;
			}
1907 1908 1909 1910
			target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16));
			kfree(p);
			break;

1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932
		default:
			printk(KERN_WARNING PFX "unknown parameter or missing value "
			       "'%s' in target creation request\n", p);
			goto out;
		}
	}

	if ((opt_mask & SRP_OPT_ALL) == SRP_OPT_ALL)
		ret = 0;
	else
		for (i = 0; i < ARRAY_SIZE(srp_opt_tokens); ++i)
			if ((srp_opt_tokens[i].token & SRP_OPT_ALL) &&
			    !(srp_opt_tokens[i].token & opt_mask))
				printk(KERN_WARNING PFX "target creation request is "
				       "missing parameter '%s'\n",
				       srp_opt_tokens[i].pattern);

out:
	kfree(options);
	return ret;
}

1933 1934
static ssize_t srp_create_target(struct device *dev,
				 struct device_attribute *attr,
1935 1936 1937
				 const char *buf, size_t count)
{
	struct srp_host *host =
1938
		container_of(dev, struct srp_host, dev);
1939 1940 1941 1942 1943 1944 1945 1946 1947 1948
	struct Scsi_Host *target_host;
	struct srp_target_port *target;
	int ret;
	int i;

	target_host = scsi_host_alloc(&srp_template,
				      sizeof (struct srp_target_port));
	if (!target_host)
		return -ENOMEM;

1949
	target_host->transportt = ib_srp_transport_template;
A
Arne Redlich 已提交
1950 1951
	target_host->max_lun     = SRP_MAX_LUN;
	target_host->max_cmd_len = sizeof ((struct srp_cmd *) (void *) 0L)->cdb;
R
Roland Dreier 已提交
1952

1953 1954
	target = host_to_target(target_host);

1955
	target->io_class   = SRP_REV16A_IB_IO_CLASS;
1956 1957
	target->scsi_host  = target_host;
	target->srp_host   = host;
1958 1959
	target->lkey	   = host->srp_dev->mr->lkey;
	target->rkey	   = host->srp_dev->mr->rkey;
1960

1961
	spin_lock_init(&target->lock);
1962
	INIT_LIST_HEAD(&target->free_tx);
1963
	INIT_LIST_HEAD(&target->free_reqs);
1964
	for (i = 0; i < SRP_CMD_SQ_SIZE; ++i) {
1965 1966 1967
		target->req_ring[i].index = i;
		list_add_tail(&target->req_ring[i].list, &target->free_reqs);
	}
1968 1969 1970 1971 1972

	ret = srp_parse_options(buf, target);
	if (ret)
		goto err;

1973
	ib_query_gid(host->srp_dev->dev, host->port, 0, &target->path.sgid);
1974

1975 1976
	shost_printk(KERN_DEBUG, target->scsi_host, PFX
		     "new target: id_ext %016llx ioc_guid %016llx pkey %04x "
H
Harvey Harrison 已提交
1977
		     "service_id %016llx dgid %pI6\n",
1978 1979 1980 1981
	       (unsigned long long) be64_to_cpu(target->id_ext),
	       (unsigned long long) be64_to_cpu(target->ioc_guid),
	       be16_to_cpu(target->path.pkey),
	       (unsigned long long) be64_to_cpu(target->service_id),
1982
	       target->path.dgid.raw);
1983 1984 1985 1986 1987

	ret = srp_create_target_ib(target);
	if (ret)
		goto err;

D
David Dillow 已提交
1988 1989
	ret = srp_new_cm_id(target);
	if (ret)
1990 1991
		goto err_free;

1992
	target->qp_in_error = 0;
1993 1994
	ret = srp_connect_target(target);
	if (ret) {
1995 1996
		shost_printk(KERN_ERR, target->scsi_host,
			     PFX "Connection failed\n");
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
		goto err_cm_id;
	}

	ret = srp_add_target(host, target);
	if (ret)
		goto err_disconnect;

	return count;

err_disconnect:
	srp_disconnect_target(target);

err_cm_id:
	ib_destroy_cm_id(target->cm_id);

err_free:
	srp_free_target_ib(target);

err:
	scsi_host_put(target_host);

	return ret;
}

2021
static DEVICE_ATTR(add_target, S_IWUSR, NULL, srp_create_target);
2022

2023 2024
static ssize_t show_ibdev(struct device *dev, struct device_attribute *attr,
			  char *buf)
2025
{
2026
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2027

2028
	return sprintf(buf, "%s\n", host->srp_dev->dev->name);
2029 2030
}

2031
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL);
2032

2033 2034
static ssize_t show_port(struct device *dev, struct device_attribute *attr,
			 char *buf)
2035
{
2036
	struct srp_host *host = container_of(dev, struct srp_host, dev);
2037 2038 2039 2040

	return sprintf(buf, "%d\n", host->port);
}

2041
static DEVICE_ATTR(port, S_IRUGO, show_port, NULL);
2042

2043
static struct srp_host *srp_add_port(struct srp_device *device, u8 port)
2044 2045 2046 2047 2048 2049 2050 2051
{
	struct srp_host *host;

	host = kzalloc(sizeof *host, GFP_KERNEL);
	if (!host)
		return NULL;

	INIT_LIST_HEAD(&host->target_list);
2052
	spin_lock_init(&host->target_lock);
2053
	init_completion(&host->released);
2054
	host->srp_dev = device;
2055 2056
	host->port = port;

2057 2058
	host->dev.class = &srp_class;
	host->dev.parent = device->dev->dma_device;
2059
	dev_set_name(&host->dev, "srp-%s-%d", device->dev->name, port);
2060

2061
	if (device_register(&host->dev))
2062
		goto free_host;
2063
	if (device_create_file(&host->dev, &dev_attr_add_target))
2064
		goto err_class;
2065
	if (device_create_file(&host->dev, &dev_attr_ibdev))
2066
		goto err_class;
2067
	if (device_create_file(&host->dev, &dev_attr_port))
2068 2069 2070 2071 2072
		goto err_class;

	return host;

err_class:
2073
	device_unregister(&host->dev);
2074

2075
free_host:
2076 2077 2078 2079 2080 2081 2082
	kfree(host);

	return NULL;
}

static void srp_add_one(struct ib_device *device)
{
2083 2084 2085
	struct srp_device *srp_dev;
	struct ib_device_attr *dev_attr;
	struct ib_fmr_pool_param fmr_param;
2086 2087 2088
	struct srp_host *host;
	int s, e, p;

2089 2090
	dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL);
	if (!dev_attr)
2091
		return;
2092

2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109
	if (ib_query_device(device, dev_attr)) {
		printk(KERN_WARNING PFX "Query device failed for %s\n",
		       device->name);
		goto free_attr;
	}

	srp_dev = kmalloc(sizeof *srp_dev, GFP_KERNEL);
	if (!srp_dev)
		goto free_attr;

	/*
	 * Use the smallest page size supported by the HCA, down to a
	 * minimum of 512 bytes (which is the smallest sector that a
	 * SCSI command will ever carry).
	 */
	srp_dev->fmr_page_shift = max(9, ffs(dev_attr->page_size_cap) - 1);
	srp_dev->fmr_page_size  = 1 << srp_dev->fmr_page_shift;
2110
	srp_dev->fmr_page_mask  = ~((u64) srp_dev->fmr_page_size - 1);
2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138

	INIT_LIST_HEAD(&srp_dev->dev_list);

	srp_dev->dev = device;
	srp_dev->pd  = ib_alloc_pd(device);
	if (IS_ERR(srp_dev->pd))
		goto free_dev;

	srp_dev->mr = ib_get_dma_mr(srp_dev->pd,
				    IB_ACCESS_LOCAL_WRITE |
				    IB_ACCESS_REMOTE_READ |
				    IB_ACCESS_REMOTE_WRITE);
	if (IS_ERR(srp_dev->mr))
		goto err_pd;

	memset(&fmr_param, 0, sizeof fmr_param);
	fmr_param.pool_size	    = SRP_FMR_POOL_SIZE;
	fmr_param.dirty_watermark   = SRP_FMR_DIRTY_SIZE;
	fmr_param.cache		    = 1;
	fmr_param.max_pages_per_fmr = SRP_FMR_SIZE;
	fmr_param.page_shift	    = srp_dev->fmr_page_shift;
	fmr_param.access	    = (IB_ACCESS_LOCAL_WRITE |
				       IB_ACCESS_REMOTE_WRITE |
				       IB_ACCESS_REMOTE_READ);

	srp_dev->fmr_pool = ib_create_fmr_pool(srp_dev->pd, &fmr_param);
	if (IS_ERR(srp_dev->fmr_pool))
		srp_dev->fmr_pool = NULL;
2139

T
Tom Tucker 已提交
2140
	if (device->node_type == RDMA_NODE_IB_SWITCH) {
2141 2142 2143 2144 2145 2146 2147 2148
		s = 0;
		e = 0;
	} else {
		s = 1;
		e = device->phys_port_cnt;
	}

	for (p = s; p <= e; ++p) {
2149
		host = srp_add_port(srp_dev, p);
2150
		if (host)
2151
			list_add_tail(&host->list, &srp_dev->dev_list);
2152 2153
	}

2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165
	ib_set_client_data(device, &srp_client, srp_dev);

	goto free_attr;

err_pd:
	ib_dealloc_pd(srp_dev->pd);

free_dev:
	kfree(srp_dev);

free_attr:
	kfree(dev_attr);
2166 2167 2168 2169
}

static void srp_remove_one(struct ib_device *device)
{
2170
	struct srp_device *srp_dev;
2171 2172 2173 2174
	struct srp_host *host, *tmp_host;
	LIST_HEAD(target_list);
	struct srp_target_port *target, *tmp_target;

2175
	srp_dev = ib_get_client_data(device, &srp_client);
2176

2177
	list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) {
2178
		device_unregister(&host->dev);
2179 2180 2181 2182 2183 2184 2185 2186 2187 2188
		/*
		 * Wait for the sysfs entry to go away, so that no new
		 * target ports can be created.
		 */
		wait_for_completion(&host->released);

		/*
		 * Mark all target ports as removed, so we stop queueing
		 * commands and don't try to reconnect.
		 */
2189
		spin_lock(&host->target_lock);
2190
		list_for_each_entry(target, &host->target_list, list) {
2191
			spin_lock_irq(&target->lock);
2192
			target->state = SRP_TARGET_REMOVED;
2193
			spin_unlock_irq(&target->lock);
2194
		}
2195
		spin_unlock(&host->target_lock);
2196 2197 2198 2199 2200 2201 2202 2203 2204 2205

		/*
		 * Wait for any reconnection tasks that may have
		 * started before we marked our target ports as
		 * removed, and any target port removal tasks.
		 */
		flush_scheduled_work();

		list_for_each_entry_safe(target, tmp_target,
					 &host->target_list, list) {
2206
			srp_remove_host(target->scsi_host);
2207
			scsi_remove_host(target->scsi_host);
2208 2209 2210 2211 2212 2213 2214 2215 2216
			srp_disconnect_target(target);
			ib_destroy_cm_id(target->cm_id);
			srp_free_target_ib(target);
			scsi_host_put(target->scsi_host);
		}

		kfree(host);
	}

2217 2218 2219 2220 2221 2222
	if (srp_dev->fmr_pool)
		ib_destroy_fmr_pool(srp_dev->fmr_pool);
	ib_dereg_mr(srp_dev->mr);
	ib_dealloc_pd(srp_dev->pd);

	kfree(srp_dev);
2223 2224
}

2225 2226 2227
static struct srp_function_template ib_srp_transport_functions = {
};

2228 2229 2230 2231
static int __init srp_init_module(void)
{
	int ret;

2232
	BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *));
2233

2234 2235 2236 2237 2238
	if (srp_sg_tablesize > 255) {
		printk(KERN_WARNING PFX "Clamping srp_sg_tablesize to 255\n");
		srp_sg_tablesize = 255;
	}

2239 2240 2241 2242 2243
	ib_srp_transport_template =
		srp_attach_transport(&ib_srp_transport_functions);
	if (!ib_srp_transport_template)
		return -ENOMEM;

2244 2245 2246 2247 2248
	srp_template.sg_tablesize = srp_sg_tablesize;
	srp_max_iu_len = (sizeof (struct srp_cmd) +
			  sizeof (struct srp_indirect_buf) +
			  srp_sg_tablesize * 16);

2249 2250 2251
	ret = class_register(&srp_class);
	if (ret) {
		printk(KERN_ERR PFX "couldn't register class infiniband_srp\n");
2252
		srp_release_transport(ib_srp_transport_template);
2253 2254 2255
		return ret;
	}

2256 2257
	ib_sa_register_client(&srp_sa_client);

2258 2259 2260
	ret = ib_register_client(&srp_client);
	if (ret) {
		printk(KERN_ERR PFX "couldn't register IB client\n");
2261
		srp_release_transport(ib_srp_transport_template);
2262
		ib_sa_unregister_client(&srp_sa_client);
2263 2264 2265 2266 2267 2268 2269 2270 2271 2272
		class_unregister(&srp_class);
		return ret;
	}

	return 0;
}

static void __exit srp_cleanup_module(void)
{
	ib_unregister_client(&srp_client);
2273
	ib_sa_unregister_client(&srp_sa_client);
2274
	class_unregister(&srp_class);
2275
	srp_release_transport(ib_srp_transport_template);
2276 2277 2278 2279
}

module_init(srp_init_module);
module_exit(srp_cleanup_module);